Accuracy of physical, statistical and neural network based algorithms for the retrieval of atmospheric water by ground-based microwave radiometry

This paper considers the estimation of atmospheric integrated precipitable water vapor and integrated liquid water content with reference to a dual-channel radiometer, including also the possible addition of a third higher frequency channel. The paper compares the performances of several retrieval algorithms applied to synthetic data of two Italian locations (Rome and Udine) on the basis of accuracy, robustness to measurement errors and applicability to sites having different climatologies from the training site.